[PATCH V5 0/5] vDPA/ifcvf: implement shared IRQ feature

[PATCH V5 0/5] vDPA/ifcvf: implement shared IRQ feature

[Zhu Lingshan]

It has been observed that on some platforms/devices, there may
not be enough MSI vectors for virtqueues and the config change.
Under such circumstances, the interrupt sources of a device
have to share vectors/IRQs.

This series implemented a shared IRQ feature for ifcvf.

Please help review.

For the test report, plase kindly refer to V3  cover letter:
https://www.spinics.net/lists/netdev/msg795808.html

Changes from V4:
(1) Insted of checking errors by themself, set_vq_vector() and
set_config_vector() would return the value read from the corresponding
registers, then let the caller do error handling.
So hw_disable() can use them.(Jason)
(2)Use virtio pci modern IO helers than ifcvf_read/write_xxx (Jason)
(3)Free all IRQs and vectors once failed in request_irq and its
subroutines, don't wait for a restart caused by start_datapath fail or
device reset issued from the guest.
(4) Better naming: vqs-reused-irq --> vqs_reused_irq,
ifcvf_dev_shared_irq --> ifcvf_dev_irq (Michael)
(5) A separate patch for ifcvf_hw cacheline alignment(Michael)

Changes from V3:
fix code indenting issues reported by LKP:

Changes from V2:
(1) Fix misuse of nvectors(in ifcvf_alloc_vectors return value)(Michael)
(2) Fix misuse of irq = get_vq_irq() in setup irqbypass(Michael)
(3) Coding style improvements(Michael)
(4) Better naming of device shared irq/shared vq irq

Changes from V1:
(1) Enable config interrupt when only one vector is allocated(Michael)
(2) Clean vectors/IRQs if failed to request config interrupt
since config interrupt is a must(Michael)
(3) Keep local vdpa_ops, disable irq_bypass by setting IRQ = -EINVAL
for shared IRQ case(Michael)
(4) Improvements on error messages(Michael)
(5) Squash functions implementation patches to the callers(Michael)

Zhu Lingshan (5):
  vDPA/ifcvf: make use of virtio pci modern IO helpers in ifcvf
  vhost_vdpa: don't setup irq offloading when irq_num < 0
  vDPA/ifcvf: implement device MSIX vector allocator
  vDPA/ifcvf: implement shared IRQ feature
  vDPA/ifcvf: cacheline alignment for ifcvf_hw

 drivers/vdpa/ifcvf/ifcvf_base.c | 120 +++++-------
 drivers/vdpa/ifcvf/ifcvf_base.h |  24 ++-
 drivers/vdpa/ifcvf/ifcvf_main.c | 317 ++++++++++++++++++++++++++++----
 drivers/vhost/vdpa.c            |   5 +-
 4 files changed, 349 insertions(+), 117 deletions(-)

-- 
2.27.0

[PATCH V5 1/5] vDPA/ifcvf: make use of virtio pci modern IO helpers in ifcvf

[Zhu Lingshan]

This commit discards ifcvf_ioreadX()/writeX(), use virtio pci
modern IO helpers instead

Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
---
 drivers/vdpa/ifcvf/ifcvf_base.c | 104 +++++++++++---------------------
 drivers/vdpa/ifcvf/ifcvf_base.h |   1 +
 drivers/vdpa/ifcvf/ifcvf_main.c |   2 +-
 3 files changed, 36 insertions(+), 71 deletions(-)

diff --git a/drivers/vdpa/ifcvf/ifcvf_base.c b/drivers/vdpa/ifcvf/ifcvf_base.c
index 7d41dfe48ade..b9fdc5258611 100644
--- a/drivers/vdpa/ifcvf/ifcvf_base.c
+++ b/drivers/vdpa/ifcvf/ifcvf_base.c
@@ -10,42 +10,6 @@
 
 #include "ifcvf_base.h"
 
-static inline u8 ifc_ioread8(u8 __iomem *addr)
-{
-	return ioread8(addr);
-}
-static inline u16 ifc_ioread16 (__le16 __iomem *addr)
-{
-	return ioread16(addr);
-}
-
-static inline u32 ifc_ioread32(__le32 __iomem *addr)
-{
-	return ioread32(addr);
-}
-
-static inline void ifc_iowrite8(u8 value, u8 __iomem *addr)
-{
-	iowrite8(value, addr);
-}
-
-static inline void ifc_iowrite16(u16 value, __le16 __iomem *addr)
-{
-	iowrite16(value, addr);
-}
-
-static inline void ifc_iowrite32(u32 value, __le32 __iomem *addr)
-{
-	iowrite32(value, addr);
-}
-
-static void ifc_iowrite64_twopart(u64 val,
-				  __le32 __iomem *lo, __le32 __iomem *hi)
-{
-	ifc_iowrite32((u32)val, lo);
-	ifc_iowrite32(val >> 32, hi);
-}
-
 struct ifcvf_adapter *vf_to_adapter(struct ifcvf_hw *hw)
 {
 	return container_of(hw, struct ifcvf_adapter, vf);
@@ -158,11 +122,11 @@ int ifcvf_init_hw(struct ifcvf_hw *hw, struct pci_dev *pdev)
 		return -EIO;
 	}
 
-	hw->nr_vring = ifc_ioread16(&hw->common_cfg->num_queues);
+	hw->nr_vring = vp_ioread16(&hw->common_cfg->num_queues);
 
 	for (i = 0; i < hw->nr_vring; i++) {
-		ifc_iowrite16(i, &hw->common_cfg->queue_select);
-		notify_off = ifc_ioread16(&hw->common_cfg->queue_notify_off);
+		vp_iowrite16(i, &hw->common_cfg->queue_select);
+		notify_off = vp_ioread16(&hw->common_cfg->queue_notify_off);
 		hw->vring[i].notify_addr = hw->notify_base +
 			notify_off * hw->notify_off_multiplier;
 		hw->vring[i].notify_pa = hw->notify_base_pa +
@@ -181,12 +145,12 @@ int ifcvf_init_hw(struct ifcvf_hw *hw, struct pci_dev *pdev)
 
 u8 ifcvf_get_status(struct ifcvf_hw *hw)
 {
-	return ifc_ioread8(&hw->common_cfg->device_status);
+	return vp_ioread8(&hw->common_cfg->device_status);
 }
 
 void ifcvf_set_status(struct ifcvf_hw *hw, u8 status)
 {
-	ifc_iowrite8(status, &hw->common_cfg->device_status);
+	vp_iowrite8(status, &hw->common_cfg->device_status);
 }
 
 void ifcvf_reset(struct ifcvf_hw *hw)
@@ -214,11 +178,11 @@ u64 ifcvf_get_hw_features(struct ifcvf_hw *hw)
 	u32 features_lo, features_hi;
 	u64 features;
 
-	ifc_iowrite32(0, &cfg->device_feature_select);
-	features_lo = ifc_ioread32(&cfg->device_feature);
+	vp_iowrite32(0, &cfg->device_feature_select);
+	features_lo = vp_ioread32(&cfg->device_feature);
 
-	ifc_iowrite32(1, &cfg->device_feature_select);
-	features_hi = ifc_ioread32(&cfg->device_feature);
+	vp_iowrite32(1, &cfg->device_feature_select);
+	features_hi = vp_ioread32(&cfg->device_feature);
 
 	features = ((u64)features_hi << 32) | features_lo;
 
@@ -271,12 +235,12 @@ void ifcvf_read_dev_config(struct ifcvf_hw *hw, u64 offset,
 
 	WARN_ON(offset + length > hw->config_size);
 	do {
-		old_gen = ifc_ioread8(&hw->common_cfg->config_generation);
+		old_gen = vp_ioread8(&hw->common_cfg->config_generation);
 		p = dst;
 		for (i = 0; i < length; i++)
-			*p++ = ifc_ioread8(hw->dev_cfg + offset + i);
+			*p++ = vp_ioread8(hw->dev_cfg + offset + i);
 
-		new_gen = ifc_ioread8(&hw->common_cfg->config_generation);
+		new_gen = vp_ioread8(&hw->common_cfg->config_generation);
 	} while (old_gen != new_gen);
 }
 
@@ -289,18 +253,18 @@ void ifcvf_write_dev_config(struct ifcvf_hw *hw, u64 offset,
 	p = src;
 	WARN_ON(offset + length > hw->config_size);
 	for (i = 0; i < length; i++)
-		ifc_iowrite8(*p++, hw->dev_cfg + offset + i);
+		vp_iowrite8(*p++, hw->dev_cfg + offset + i);
 }
 
 static void ifcvf_set_features(struct ifcvf_hw *hw, u64 features)
 {
 	struct virtio_pci_common_cfg __iomem *cfg = hw->common_cfg;
 
-	ifc_iowrite32(0, &cfg->guest_feature_select);
-	ifc_iowrite32((u32)features, &cfg->guest_feature);
+	vp_iowrite32(0, &cfg->guest_feature_select);
+	vp_iowrite32((u32)features, &cfg->guest_feature);
 
-	ifc_iowrite32(1, &cfg->guest_feature_select);
-	ifc_iowrite32(features >> 32, &cfg->guest_feature);
+	vp_iowrite32(1, &cfg->guest_feature_select);
+	vp_iowrite32(features >> 32, &cfg->guest_feature);
 }
 
 static int ifcvf_config_features(struct ifcvf_hw *hw)
@@ -329,7 +293,7 @@ u16 ifcvf_get_vq_state(struct ifcvf_hw *hw, u16 qid)
 	ifcvf_lm = (struct ifcvf_lm_cfg __iomem *)hw->lm_cfg;
 	q_pair_id = qid / hw->nr_vring;
 	avail_idx_addr = &ifcvf_lm->vring_lm_cfg[q_pair_id].idx_addr[qid % 2];
-	last_avail_idx = ifc_ioread16(avail_idx_addr);
+	last_avail_idx = vp_ioread16(avail_idx_addr);
 
 	return last_avail_idx;
 }
@@ -344,7 +308,7 @@ int ifcvf_set_vq_state(struct ifcvf_hw *hw, u16 qid, u16 num)
 	q_pair_id = qid / hw->nr_vring;
 	avail_idx_addr = &ifcvf_lm->vring_lm_cfg[q_pair_id].idx_addr[qid % 2];
 	hw->vring[qid].last_avail_idx = num;
-	ifc_iowrite16(num, avail_idx_addr);
+	vp_iowrite16(num, avail_idx_addr);
 
 	return 0;
 }
@@ -357,9 +321,9 @@ static int ifcvf_hw_enable(struct ifcvf_hw *hw)
 
 	ifcvf = vf_to_adapter(hw);
 	cfg = hw->common_cfg;
-	ifc_iowrite16(IFCVF_MSI_CONFIG_OFF, &cfg->msix_config);
+	vp_iowrite16(IFCVF_MSI_CONFIG_OFF, &cfg->msix_config);
 
-	if (ifc_ioread16(&cfg->msix_config) == VIRTIO_MSI_NO_VECTOR) {
+	if (vp_ioread16(&cfg->msix_config) == VIRTIO_MSI_NO_VECTOR) {
 		IFCVF_ERR(ifcvf->pdev, "No msix vector for device config\n");
 		return -EINVAL;
 	}
@@ -368,17 +332,17 @@ static int ifcvf_hw_enable(struct ifcvf_hw *hw)
 		if (!hw->vring[i].ready)
 			break;
 
-		ifc_iowrite16(i, &cfg->queue_select);
-		ifc_iowrite64_twopart(hw->vring[i].desc, &cfg->queue_desc_lo,
+		vp_iowrite16(i, &cfg->queue_select);
+		vp_iowrite64_twopart(hw->vring[i].desc, &cfg->queue_desc_lo,
 				     &cfg->queue_desc_hi);
-		ifc_iowrite64_twopart(hw->vring[i].avail, &cfg->queue_avail_lo,
+		vp_iowrite64_twopart(hw->vring[i].avail, &cfg->queue_avail_lo,
 				      &cfg->queue_avail_hi);
-		ifc_iowrite64_twopart(hw->vring[i].used, &cfg->queue_used_lo,
+		vp_iowrite64_twopart(hw->vring[i].used, &cfg->queue_used_lo,
 				     &cfg->queue_used_hi);
-		ifc_iowrite16(hw->vring[i].size, &cfg->queue_size);
-		ifc_iowrite16(i + IFCVF_MSI_QUEUE_OFF, &cfg->queue_msix_vector);
+		vp_iowrite16(hw->vring[i].size, &cfg->queue_size);
+		vp_iowrite16(i + IFCVF_MSI_QUEUE_OFF, &cfg->queue_msix_vector);
 
-		if (ifc_ioread16(&cfg->queue_msix_vector) ==
+		if (vp_ioread16(&cfg->queue_msix_vector) ==
 		    VIRTIO_MSI_NO_VECTOR) {
 			IFCVF_ERR(ifcvf->pdev,
 				  "No msix vector for queue %u\n", i);
@@ -386,7 +350,7 @@ static int ifcvf_hw_enable(struct ifcvf_hw *hw)
 		}
 
 		ifcvf_set_vq_state(hw, i, hw->vring[i].last_avail_idx);
-		ifc_iowrite16(1, &cfg->queue_enable);
+		vp_iowrite16(1, &cfg->queue_enable);
 	}
 
 	return 0;
@@ -398,14 +362,14 @@ static void ifcvf_hw_disable(struct ifcvf_hw *hw)
 	u32 i;
 
 	cfg = hw->common_cfg;
-	ifc_iowrite16(VIRTIO_MSI_NO_VECTOR, &cfg->msix_config);
+	vp_iowrite16(VIRTIO_MSI_NO_VECTOR, &cfg->msix_config);
 
 	for (i = 0; i < hw->nr_vring; i++) {
-		ifc_iowrite16(i, &cfg->queue_select);
-		ifc_iowrite16(VIRTIO_MSI_NO_VECTOR, &cfg->queue_msix_vector);
+		vp_iowrite16(i, &cfg->queue_select);
+		vp_iowrite16(VIRTIO_MSI_NO_VECTOR, &cfg->queue_msix_vector);
 	}
 
-	ifc_ioread16(&cfg->queue_msix_vector);
+	vp_ioread16(&cfg->queue_msix_vector);
 }
 
 int ifcvf_start_hw(struct ifcvf_hw *hw)
@@ -433,5 +397,5 @@ void ifcvf_stop_hw(struct ifcvf_hw *hw)
 
 void ifcvf_notify_queue(struct ifcvf_hw *hw, u16 qid)
 {
-	ifc_iowrite16(qid, hw->vring[qid].notify_addr);
+	vp_iowrite16(qid, hw->vring[qid].notify_addr);
 }
diff --git a/drivers/vdpa/ifcvf/ifcvf_base.h b/drivers/vdpa/ifcvf/ifcvf_base.h
index c486873f370a..25c591a3eae2 100644
--- a/drivers/vdpa/ifcvf/ifcvf_base.h
+++ b/drivers/vdpa/ifcvf/ifcvf_base.h
@@ -14,6 +14,7 @@
 #include <linux/pci.h>
 #include <linux/pci_regs.h>
 #include <linux/vdpa.h>
+#include <linux/virtio_pci_modern.h>
 #include <uapi/linux/virtio_net.h>
 #include <uapi/linux/virtio_blk.h>
 #include <uapi/linux/virtio_config.h>
diff --git a/drivers/vdpa/ifcvf/ifcvf_main.c b/drivers/vdpa/ifcvf/ifcvf_main.c
index d1a6b5ab543c..43b7180256c6 100644
--- a/drivers/vdpa/ifcvf/ifcvf_main.c
+++ b/drivers/vdpa/ifcvf/ifcvf_main.c
@@ -348,7 +348,7 @@ static u32 ifcvf_vdpa_get_generation(struct vdpa_device *vdpa_dev)
 {
 	struct ifcvf_hw *vf = vdpa_to_vf(vdpa_dev);
 
-	return ioread8(&vf->common_cfg->config_generation);
+	return vp_ioread8(&vf->common_cfg->config_generation);
 }
 
 static u32 ifcvf_vdpa_get_device_id(struct vdpa_device *vdpa_dev)
-- 
2.27.0

[PATCH V5 2/5] vhost_vdpa: don't setup irq offloading when irq_num < 0

[Zhu Lingshan]

When irq number is negative(e.g., -EINVAL), the virtqueue
may be disabled or the virtqueues are sharing a device irq.
In such case, we should not setup irq offloading for a virtqueue.

Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
---
 drivers/vhost/vdpa.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drivers/vhost/vdpa.c b/drivers/vhost/vdpa.c
index 851539807bc9..8f53a8478c28 100644
--- a/drivers/vhost/vdpa.c
+++ b/drivers/vhost/vdpa.c
@@ -97,8 +97,11 @@ static void vhost_vdpa_setup_vq_irq(struct vhost_vdpa *v, u16 qid)
 		return;
 
 	irq = ops->get_vq_irq(vdpa, qid);
+	if (irq < 0)
+		return;
+
 	irq_bypass_unregister_producer(&vq->call_ctx.producer);
-	if (!vq->call_ctx.ctx || irq < 0)
+	if (!vq->call_ctx.ctx)
 		return;
 
 	vq->call_ctx.producer.token = vq->call_ctx.ctx;
-- 
2.27.0

[PATCH V5 3/5] vDPA/ifcvf: implement device MSIX vector allocator

[Zhu Lingshan]

This commit implements a MSIX vector allocation helper
for vqs and config interrupts.

Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
---
 drivers/vdpa/ifcvf/ifcvf_main.c | 31 ++++++++++++++++++++++++++-----
 1 file changed, 26 insertions(+), 5 deletions(-)

diff --git a/drivers/vdpa/ifcvf/ifcvf_main.c b/drivers/vdpa/ifcvf/ifcvf_main.c
index 43b7180256c6..964f7ac142ba 100644
--- a/drivers/vdpa/ifcvf/ifcvf_main.c
+++ b/drivers/vdpa/ifcvf/ifcvf_main.c
@@ -58,23 +58,44 @@ static void ifcvf_free_irq(struct ifcvf_adapter *adapter, int queues)
 	ifcvf_free_irq_vectors(pdev);
 }
 
-static int ifcvf_request_irq(struct ifcvf_adapter *adapter)
+/* ifcvf MSIX vectors allocator, this helper tries to allocate
+ * vectors for all virtqueues and the config interrupt.
+ * It returns the number of allocated vectors, negative
+ * return value when fails.
+ */
+static int ifcvf_alloc_vectors(struct ifcvf_adapter *adapter)
 {
 	struct pci_dev *pdev = adapter->pdev;
 	struct ifcvf_hw *vf = &adapter->vf;
-	int vector, i, ret, irq;
-	u16 max_intr;
+	int max_intr, ret;
 
 	/* all queues and config interrupt  */
 	max_intr = vf->nr_vring + 1;
+	ret = pci_alloc_irq_vectors(pdev, 1, max_intr, PCI_IRQ_MSIX | PCI_IRQ_AFFINITY);
 
-	ret = pci_alloc_irq_vectors(pdev, max_intr,
-				    max_intr, PCI_IRQ_MSIX);
 	if (ret < 0) {
 		IFCVF_ERR(pdev, "Failed to alloc IRQ vectors\n");
 		return ret;
 	}
 
+	if (ret < max_intr)
+		IFCVF_INFO(pdev,
+			   "Requested %u vectors, however only %u allocated, lower performance\n",
+			   max_intr, ret);
+
+	return ret;
+}
+
+static int ifcvf_request_irq(struct ifcvf_adapter *adapter)
+{
+	struct pci_dev *pdev = adapter->pdev;
+	struct ifcvf_hw *vf = &adapter->vf;
+	int vector, nvectors, i, ret, irq;
+
+	nvectors = ifcvf_alloc_vectors(adapter);
+	if (nvectors <= 0)
+		return -EFAULT;
+
 	snprintf(vf->config_msix_name, 256, "ifcvf[%s]-config\n",
 		 pci_name(pdev));
 	vector = 0;
-- 
2.27.0

[PATCH V5 4/5] vDPA/ifcvf: implement shared IRQ feature

[Zhu Lingshan]

On some platforms/devices, there may not be enough MSI vectors
allocated for the virtqueues and config changes. In such a case,
the interrupt sources(virtqueues, config changes) must share
an IRQ/vector, to avoid initialization failures, keep
the device functional.

This commit handles three cases:
(1) number of the allocated vectors == the number of virtqueues + 1
(config changes), every virtqueue and the config interrupt has
a separated vector/IRQ, the best and the most likely case.
(2) number of the allocated vectors is less than the best case, but
greater than 1. In this case, all virtqueues share a vector/IRQ,
the config interrupt has a separated vector/IRQ
(3) only one vector is allocated, in this case, the virtqueues and
the config interrupt share a vector/IRQ. The worst and most
unlikely case.

Otherwise, it needs to fail.

This commit introduces some helper functions:
ifcvf_set_vq_vector() and ifcvf_set_config_vector() sets virtqueue
vector and config vector in the device config space, so that
the device can send interrupt DMA.

Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
---
 drivers/vdpa/ifcvf/ifcvf_base.c |  48 +++---
 drivers/vdpa/ifcvf/ifcvf_base.h |  15 +-
 drivers/vdpa/ifcvf/ifcvf_main.c | 294 ++++++++++++++++++++++++++++----
 3 files changed, 300 insertions(+), 57 deletions(-)

diff --git a/drivers/vdpa/ifcvf/ifcvf_base.c b/drivers/vdpa/ifcvf/ifcvf_base.c
index b9fdc5258611..ba4866f871dd 100644
--- a/drivers/vdpa/ifcvf/ifcvf_base.c
+++ b/drivers/vdpa/ifcvf/ifcvf_base.c
@@ -15,6 +15,26 @@ struct ifcvf_adapter *vf_to_adapter(struct ifcvf_hw *hw)
 	return container_of(hw, struct ifcvf_adapter, vf);
 }
 
+u16 ifcvf_set_vq_vector(struct ifcvf_hw *hw, u16 qid, int vector)
+{
+	struct virtio_pci_common_cfg __iomem *cfg = hw->common_cfg;
+
+	vp_iowrite16(qid, &cfg->queue_select);
+	vp_iowrite16(vector, &cfg->queue_msix_vector);
+
+	return vp_ioread16(&cfg->queue_msix_vector);
+}
+
+u16 ifcvf_set_config_vector(struct ifcvf_hw *hw, int vector)
+{
+	struct virtio_pci_common_cfg __iomem *cfg = hw->common_cfg;
+
+	cfg = hw->common_cfg;
+	vp_iowrite16(vector,  &cfg->msix_config);
+
+	return vp_ioread16(&cfg->msix_config);
+}
+
 static void __iomem *get_cap_addr(struct ifcvf_hw *hw,
 				  struct virtio_pci_cap *cap)
 {
@@ -131,6 +151,7 @@ int ifcvf_init_hw(struct ifcvf_hw *hw, struct pci_dev *pdev)
 			notify_off * hw->notify_off_multiplier;
 		hw->vring[i].notify_pa = hw->notify_base_pa +
 			notify_off * hw->notify_off_multiplier;
+		hw->vring[i].irq = -EINVAL;
 	}
 
 	hw->lm_cfg = hw->base[IFCVF_LM_BAR];
@@ -140,6 +161,9 @@ int ifcvf_init_hw(struct ifcvf_hw *hw, struct pci_dev *pdev)
 		  hw->common_cfg, hw->notify_base, hw->isr,
 		  hw->dev_cfg, hw->notify_off_multiplier);
 
+	hw->vqs_reused_irq = -EINVAL;
+	hw->config_irq = -EINVAL;
+
 	return 0;
 }
 
@@ -321,13 +345,6 @@ static int ifcvf_hw_enable(struct ifcvf_hw *hw)
 
 	ifcvf = vf_to_adapter(hw);
 	cfg = hw->common_cfg;
-	vp_iowrite16(IFCVF_MSI_CONFIG_OFF, &cfg->msix_config);
-
-	if (vp_ioread16(&cfg->msix_config) == VIRTIO_MSI_NO_VECTOR) {
-		IFCVF_ERR(ifcvf->pdev, "No msix vector for device config\n");
-		return -EINVAL;
-	}
-
 	for (i = 0; i < hw->nr_vring; i++) {
 		if (!hw->vring[i].ready)
 			break;
@@ -340,15 +357,6 @@ static int ifcvf_hw_enable(struct ifcvf_hw *hw)
 		vp_iowrite64_twopart(hw->vring[i].used, &cfg->queue_used_lo,
 				     &cfg->queue_used_hi);
 		vp_iowrite16(hw->vring[i].size, &cfg->queue_size);
-		vp_iowrite16(i + IFCVF_MSI_QUEUE_OFF, &cfg->queue_msix_vector);
-
-		if (vp_ioread16(&cfg->queue_msix_vector) ==
-		    VIRTIO_MSI_NO_VECTOR) {
-			IFCVF_ERR(ifcvf->pdev,
-				  "No msix vector for queue %u\n", i);
-			return -EINVAL;
-		}
-
 		ifcvf_set_vq_state(hw, i, hw->vring[i].last_avail_idx);
 		vp_iowrite16(1, &cfg->queue_enable);
 	}
@@ -362,14 +370,10 @@ static void ifcvf_hw_disable(struct ifcvf_hw *hw)
 	u32 i;
 
 	cfg = hw->common_cfg;
-	vp_iowrite16(VIRTIO_MSI_NO_VECTOR, &cfg->msix_config);
-
+	ifcvf_set_config_vector(hw, VIRTIO_MSI_NO_VECTOR);
 	for (i = 0; i < hw->nr_vring; i++) {
-		vp_iowrite16(i, &cfg->queue_select);
-		vp_iowrite16(VIRTIO_MSI_NO_VECTOR, &cfg->queue_msix_vector);
+		ifcvf_set_vq_vector(hw, i, VIRTIO_MSI_NO_VECTOR)
 	}
-
-	vp_ioread16(&cfg->queue_msix_vector);
 }
 
 int ifcvf_start_hw(struct ifcvf_hw *hw)
diff --git a/drivers/vdpa/ifcvf/ifcvf_base.h b/drivers/vdpa/ifcvf/ifcvf_base.h
index 25c591a3eae2..dcd31accfce5 100644
--- a/drivers/vdpa/ifcvf/ifcvf_base.h
+++ b/drivers/vdpa/ifcvf/ifcvf_base.h
@@ -28,8 +28,6 @@
 
 #define IFCVF_QUEUE_ALIGNMENT	PAGE_SIZE
 #define IFCVF_QUEUE_MAX		32768
-#define IFCVF_MSI_CONFIG_OFF	0
-#define IFCVF_MSI_QUEUE_OFF	1
 #define IFCVF_PCI_MAX_RESOURCE	6
 
 #define IFCVF_LM_CFG_SIZE		0x40
@@ -43,6 +41,13 @@
 #define ifcvf_private_to_vf(adapter) \
 	(&((struct ifcvf_adapter *)adapter)->vf)
 
+/* all vqs and config interrupt has its own vector */
+#define MSIX_VECTOR_PER_VQ_AND_CONFIG		1
+/* all vqs share a vector, and config interrupt has a separate vector */
+#define MSIX_VECTOR_SHARED_VQ_AND_CONFIG	2
+/* all vqs and config interrupt share a vector */
+#define MSIX_VECTOR_DEV_SHARED			3
+
 struct vring_info {
 	u64 desc;
 	u64 avail;
@@ -77,9 +82,11 @@ struct ifcvf_hw {
 	void __iomem * const *base;
 	char config_msix_name[256];
 	struct vdpa_callback config_cb;
-	unsigned int config_irq;
+	int config_irq;
+	int vqs_reused_irq;
 	/* virtio-net or virtio-blk device config size */
 	u32 config_size;
+	u8 msix_vector_status;
 };
 
 struct ifcvf_adapter {
@@ -124,4 +131,6 @@ int ifcvf_set_vq_state(struct ifcvf_hw *hw, u16 qid, u16 num);
 struct ifcvf_adapter *vf_to_adapter(struct ifcvf_hw *hw);
 int ifcvf_probed_virtio_net(struct ifcvf_hw *hw);
 u32 ifcvf_get_config_size(struct ifcvf_hw *hw);
+u16 ifcvf_set_vq_vector(struct ifcvf_hw *hw, u16 qid, int vector);
+u16 ifcvf_set_config_vector(struct ifcvf_hw *hw, int vector);
 #endif /* _IFCVF_H_ */
diff --git a/drivers/vdpa/ifcvf/ifcvf_main.c b/drivers/vdpa/ifcvf/ifcvf_main.c
index 964f7ac142ba..3b48e717e89f 100644
--- a/drivers/vdpa/ifcvf/ifcvf_main.c
+++ b/drivers/vdpa/ifcvf/ifcvf_main.c
@@ -27,7 +27,7 @@ static irqreturn_t ifcvf_config_changed(int irq, void *arg)
 	return IRQ_HANDLED;
 }
 
-static irqreturn_t ifcvf_intr_handler(int irq, void *arg)
+static irqreturn_t ifcvf_vq_intr_handler(int irq, void *arg)
 {
 	struct vring_info *vring = arg;
 
@@ -37,24 +37,98 @@ static irqreturn_t ifcvf_intr_handler(int irq, void *arg)
 	return IRQ_HANDLED;
 }
 
+static irqreturn_t ifcvf_vqs_reused_intr_handler(int irq, void *arg)
+{
+	struct ifcvf_hw *vf = arg;
+	struct vring_info *vring;
+	int i;
+
+	for (i = 0; i < vf->nr_vring; i++) {
+		vring = &vf->vring[i];
+		if (vring->cb.callback)
+			vf->vring->cb.callback(vring->cb.private);
+	}
+
+	return IRQ_HANDLED;
+}
+
+static irqreturn_t ifcvf_dev_intr_handler(int irq, void *arg)
+{
+	struct ifcvf_hw *vf = arg;
+	u8 isr;
+
+	isr = vp_ioread8(vf->isr);
+	if (isr & VIRTIO_PCI_ISR_CONFIG)
+		ifcvf_config_changed(irq, arg);
+
+	return ifcvf_vqs_reused_intr_handler(irq, arg);
+}
+
 static void ifcvf_free_irq_vectors(void *data)
 {
 	pci_free_irq_vectors(data);
 }
 
-static void ifcvf_free_irq(struct ifcvf_adapter *adapter, int queues)
+static void ifcvf_free_per_vq_irq(struct ifcvf_adapter *adapter)
 {
 	struct pci_dev *pdev = adapter->pdev;
 	struct ifcvf_hw *vf = &adapter->vf;
 	int i;
 
+	for (i = 0; i < vf->nr_vring; i++) {
+		if (vf->vring[i].irq != -EINVAL) {
+			devm_free_irq(&pdev->dev, vf->vring[i].irq, &vf->vring[i]);
+			vf->vring[i].irq = -EINVAL;
+		}
+	}
+}
 
-	for (i = 0; i < queues; i++) {
-		devm_free_irq(&pdev->dev, vf->vring[i].irq, &vf->vring[i]);
-		vf->vring[i].irq = -EINVAL;
+static void ifcvf_free_vqs_reused_irq(struct ifcvf_adapter *adapter)
+{
+	struct pci_dev *pdev = adapter->pdev;
+	struct ifcvf_hw *vf = &adapter->vf;
+
+	if (vf->vqs_reused_irq != -EINVAL) {
+		devm_free_irq(&pdev->dev, vf->vqs_reused_irq, vf);
+		vf->vqs_reused_irq = -EINVAL;
 	}
 
-	devm_free_irq(&pdev->dev, vf->config_irq, vf);
+}
+
+static void ifcvf_free_vq_irq(struct ifcvf_adapter *adapter)
+{
+	struct ifcvf_hw *vf = &adapter->vf;
+
+	if (vf->msix_vector_status == MSIX_VECTOR_PER_VQ_AND_CONFIG)
+		ifcvf_free_per_vq_irq(adapter);
+	else
+		ifcvf_free_vqs_reused_irq(adapter);
+}
+
+static void ifcvf_free_config_irq(struct ifcvf_adapter *adapter)
+{
+	struct pci_dev *pdev = adapter->pdev;
+	struct ifcvf_hw *vf = &adapter->vf;
+
+	if (vf->config_irq == -EINVAL)
+		return;
+
+	/* If the irq is shared by all vqs and the config interrupt,
+	 * it is already freed in ifcvf_free_vq_irq, so here only
+	 * need to free config irq when msix_vector_status != MSIX_VECTOR_DEV_SHARED
+	 */
+	if (vf->msix_vector_status != MSIX_VECTOR_DEV_SHARED) {
+		devm_free_irq(&pdev->dev, vf->config_irq, vf);
+		vf->config_irq = -EINVAL;
+	}
+}
+
+static void ifcvf_free_irq(struct ifcvf_adapter *adapter)
+{
+	struct pci_dev *pdev = adapter->pdev;
+
+	ifcvf_free_vq_irq(adapter);
+	ifcvf_free_config_irq(adapter);
 	ifcvf_free_irq_vectors(pdev);
 }
 
@@ -86,48 +160,201 @@ static int ifcvf_alloc_vectors(struct ifcvf_adapter *adapter)
 	return ret;
 }
 
-static int ifcvf_request_irq(struct ifcvf_adapter *adapter)
+static int ifcvf_request_per_vq_irq(struct ifcvf_adapter *adapter)
 {
 	struct pci_dev *pdev = adapter->pdev;
 	struct ifcvf_hw *vf = &adapter->vf;
-	int vector, nvectors, i, ret, irq;
+	int i, vector, ret, irq;
 
-	nvectors = ifcvf_alloc_vectors(adapter);
-	if (nvectors <= 0)
-		return -EFAULT;
+	vf->vqs_reused_irq = -EINVAL;
+	for (i = 0; i < vf->nr_vring; i++) {
+		snprintf(vf->vring[i].msix_name, 256, "ifcvf[%s]-%d\n", pci_name(pdev), i);
+		vector = i;
+		irq = pci_irq_vector(pdev, vector);
+		ret = devm_request_irq(&pdev->dev, irq,
+				       ifcvf_vq_intr_handler, 0,
+				       vf->vring[i].msix_name,
+				       &vf->vring[i]);
+		if (ret) {
+			IFCVF_ERR(pdev, "Failed to request irq for vq %d\n", i);
+			goto err;
+		}
+
+		vf->vring[i].irq = irq;
+		ret = ifcvf_set_vq_vector(vf, i, vector);
+		if (ret == VIRTIO_MSI_NO_VECTOR) {
+			IFCVF_ERR(pdev, "No msix vector for vq %u\n", i);
+			goto err;
+		}
+	}
+
+	return 0;
+err:
+	ifcvf_free_irq(adapter);
+
+	return -EFAULT;
+}
+
+static int ifcvf_request_vqs_reused_irq(struct ifcvf_adapter *adapter)
+{
+	struct pci_dev *pdev = adapter->pdev;
+	struct ifcvf_hw *vf = &adapter->vf;
+	int i, vector, ret, irq;
+
+	vector = 0;
+	snprintf(vf->vring[0].msix_name, 256, "ifcvf[%s]-vqs-reused-irq\n", pci_name(pdev));
+	irq = pci_irq_vector(pdev, vector);
+	ret = devm_request_irq(&pdev->dev, irq,
+			       ifcvf_vqs_reused_intr_handler, 0,
+			       vf->vring[0].msix_name, vf);
+	if (ret) {
+		IFCVF_ERR(pdev, "Failed to request reused irq for the device\n");
+		goto err;
+	}
+
+	vf->vqs_reused_irq = irq;
+	for (i = 0; i < vf->nr_vring; i++) {
+		vf->vring[i].irq = -EINVAL;
+		ret = ifcvf_set_vq_vector(vf, i, vector);
+		if (ret == VIRTIO_MSI_NO_VECTOR) {
+			IFCVF_ERR(pdev, "No msix vector for vq %u\n", i);
+			goto err;
+		}
+	}
+
+	return 0;
+err:
+	ifcvf_free_irq(adapter);
+
+	return -EFAULT;
+}
+
+static int ifcvf_request_dev_irq(struct ifcvf_adapter *adapter)
+{
+	struct pci_dev *pdev = adapter->pdev;
+	struct ifcvf_hw *vf = &adapter->vf;
+	int i, vector, ret, irq;
+
+	vector = 0;
+	snprintf(vf->vring[0].msix_name, 256, "ifcvf[%s]-dev-irq\n", pci_name(pdev));
+	irq = pci_irq_vector(pdev, vector);
+	ret = devm_request_irq(&pdev->dev, irq,
+			       ifcvf_dev_intr_handler, 0,
+			       vf->vring[0].msix_name, vf);
+	if (ret) {
+		IFCVF_ERR(pdev, "Failed to request irq for the device\n");
+		goto err;
+	}
+
+	vf->vqs_reused_irq = irq;
+	for (i = 0; i < vf->nr_vring; i++) {
+		vf->vring[i].irq = -EINVAL;
+		ret = ifcvf_set_vq_vector(vf, i, vector);
+		if (ret == VIRTIO_MSI_NO_VECTOR) {
+			IFCVF_ERR(pdev, "No msix vector for vq %u\n", i);
+			goto err;
+		}
+	}
+
+	vf->config_irq = irq;
+	ret = ifcvf_set_config_vector(vf, vector);
+	if (ret == VIRTIO_MSI_NO_VECTOR) {
+		IFCVF_ERR(pdev, "No msix vector for device config\n");
+		goto err;
+	}
+
+	return 0;
+err:
+	ifcvf_free_irq(adapter);
+
+	return -EFAULT;
+
+}
+
+static int ifcvf_request_vq_irq(struct ifcvf_adapter *adapter)
+{
+	struct ifcvf_hw *vf = &adapter->vf;
+	int ret;
+
+	if (vf->msix_vector_status == MSIX_VECTOR_PER_VQ_AND_CONFIG)
+		ret = ifcvf_request_per_vq_irq(adapter);
+	else
+		ret = ifcvf_request_vqs_reused_irq(adapter);
+
+	return ret;
+}
+
+static int ifcvf_request_config_irq(struct ifcvf_adapter *adapter)
+{
+	struct pci_dev *pdev = adapter->pdev;
+	struct ifcvf_hw *vf = &adapter->vf;
+	int config_vector, ret;
+
+	if (vf->msix_vector_status == MSIX_VECTOR_DEV_SHARED)
+		return 0;
+
+	if (vf->msix_vector_status == MSIX_VECTOR_PER_VQ_AND_CONFIG)
+		/* vector 0 ~ vf->nr_vring for vqs, num vf->nr_vring vector for config interrupt */
+		config_vector = vf->nr_vring;
+
+	if (vf->msix_vector_status ==  MSIX_VECTOR_SHARED_VQ_AND_CONFIG)
+		/* vector 0 for vqs and 1 for config interrupt */
+		config_vector = 1;
 
 	snprintf(vf->config_msix_name, 256, "ifcvf[%s]-config\n",
 		 pci_name(pdev));
-	vector = 0;
-	vf->config_irq = pci_irq_vector(pdev, vector);
+	vf->config_irq = pci_irq_vector(pdev, config_vector);
 	ret = devm_request_irq(&pdev->dev, vf->config_irq,
 			       ifcvf_config_changed, 0,
 			       vf->config_msix_name, vf);
 	if (ret) {
 		IFCVF_ERR(pdev, "Failed to request config irq\n");
-		return ret;
+		goto err;
 	}
 
-	for (i = 0; i < vf->nr_vring; i++) {
-		snprintf(vf->vring[i].msix_name, 256, "ifcvf[%s]-%d\n",
-			 pci_name(pdev), i);
-		vector = i + IFCVF_MSI_QUEUE_OFF;
-		irq = pci_irq_vector(pdev, vector);
-		ret = devm_request_irq(&pdev->dev, irq,
-				       ifcvf_intr_handler, 0,
-				       vf->vring[i].msix_name,
-				       &vf->vring[i]);
-		if (ret) {
-			IFCVF_ERR(pdev,
-				  "Failed to request irq for vq %d\n", i);
-			ifcvf_free_irq(adapter, i);
+	ret = ifcvf_set_config_vector(vf, config_vector);
+	if (ret == VIRTIO_MSI_NO_VECTOR) {
+		IFCVF_ERR(pdev, "No msix vector for device config\n");
+		goto err;
+	}
 
-			return ret;
-		}
+	return 0;
+err:
+	ifcvf_free_irq(adapter);
 
-		vf->vring[i].irq = irq;
+	return -EFAULT;
+}
+
+static int ifcvf_request_irq(struct ifcvf_adapter *adapter)
+{
+	struct ifcvf_hw *vf = &adapter->vf;
+	int nvectors, ret, max_intr;
+
+	nvectors = ifcvf_alloc_vectors(adapter);
+	if (nvectors <= 0)
+		return -EFAULT;
+
+	vf->msix_vector_status = MSIX_VECTOR_PER_VQ_AND_CONFIG;
+	max_intr = vf->nr_vring + 1;
+	if (nvectors < max_intr)
+		vf->msix_vector_status = MSIX_VECTOR_SHARED_VQ_AND_CONFIG;
+
+	if (nvectors == 1) {
+		vf->msix_vector_status = MSIX_VECTOR_DEV_SHARED;
+		ret = ifcvf_request_dev_irq(adapter);
+
+		return ret;
 	}
 
+	ret = ifcvf_request_vq_irq(adapter);
+	if (ret)
+		return ret;
+
+	ret = ifcvf_request_config_irq(adapter);
+
+	if (ret)
+		return ret;
+
 	return 0;
 }
 
@@ -284,7 +511,7 @@ static int ifcvf_vdpa_reset(struct vdpa_device *vdpa_dev)
 
 	if (status_old & VIRTIO_CONFIG_S_DRIVER_OK) {
 		ifcvf_stop_datapath(adapter);
-		ifcvf_free_irq(adapter, vf->nr_vring);
+		ifcvf_free_irq(adapter);
 	}
 
 	ifcvf_reset_vring(adapter);
@@ -431,7 +658,10 @@ static int ifcvf_vdpa_get_vq_irq(struct vdpa_device *vdpa_dev,
 {
 	struct ifcvf_hw *vf = vdpa_to_vf(vdpa_dev);
 
-	return vf->vring[qid].irq;
+	if (vf->vqs_reused_irq < 0)
+		return vf->vring[qid].irq;
+	else
+		return -EINVAL;
 }
 
 static struct vdpa_notification_area ifcvf_get_vq_notification(struct vdpa_device *vdpa_dev,
-- 
2.27.0

[PATCH V5 5/5] vDPA/ifcvf: cacheline alignment for ifcvf_hw

[Zhu Lingshan]

This commit introduces a new cacheline aligned layout for
ifcvf_hw.

Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
---
 drivers/vdpa/ifcvf/ifcvf_base.h | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/drivers/vdpa/ifcvf/ifcvf_base.h b/drivers/vdpa/ifcvf/ifcvf_base.h
index dcd31accfce5..115b61f4924b 100644
--- a/drivers/vdpa/ifcvf/ifcvf_base.h
+++ b/drivers/vdpa/ifcvf/ifcvf_base.h
@@ -66,16 +66,18 @@ struct ifcvf_hw {
 	u8 __iomem *isr;
 	/* Live migration */
 	u8 __iomem *lm_cfg;
-	u16 nr_vring;
 	/* Notification bar number */
 	u8 notify_bar;
+	u8 msix_vector_status;
+	/* virtio-net or virtio-blk device config size */
+	u32 config_size;
 	/* Notificaiton bar address */
 	void __iomem *notify_base;
 	phys_addr_t notify_base_pa;
 	u32 notify_off_multiplier;
+	u32 dev_type;
 	u64 req_features;
 	u64 hw_features;
-	u32 dev_type;
 	struct virtio_pci_common_cfg __iomem *common_cfg;
 	void __iomem *dev_cfg;
 	struct vring_info vring[IFCVF_MAX_QUEUES];
@@ -84,9 +86,7 @@ struct ifcvf_hw {
 	struct vdpa_callback config_cb;
 	int config_irq;
 	int vqs_reused_irq;
-	/* virtio-net or virtio-blk device config size */
-	u32 config_size;
-	u8 msix_vector_status;
+	u16 nr_vring;
 };
 
 struct ifcvf_adapter {
-- 
2.27.0